DeepMind AI showing signs of human intuition -- improvising like a human during chess matches in "turning point"

What i find intriguing is that there was an embargo on the game data for its games against stockfish. A limited number were released then a year later a limited number more after an embargo..
It was literally so impressive they made its chess games SECRET.

Also if i was in charge of running the experiment at deep mind, i would have left Alpha Zero training all year not just a week or so.. Even though it was law of diminishing returns. I want to know if it somehow solved chess, and just how high of a rating it could achieve.

Also what would happen if it played a brute force chess engine running on some serious computer hardware like the Summit supercomputer capable of 1.88 exaops during a genomic analysis and is expected to reach 3.3 exaops using mixed precision calculations. 143 Petaflops double precision 64bit. 2.8 million gigs of Ram.

Last Edited by Agent Smith 2014 on 12/08/2018 03:39 PM

So they're trying not to scare the shit out of the public because the AI is now so advanced?

Reminds me of Elon Musk saying "We are building progressively greater intelligence, and the percentage of intelligence that is not human is increasing and eventually we will represent a very small percentage of intelligence."


[link to www.youtube.com (secure)]

Last Edited by STAX on 12/08/2018 03:44 PM

yeah if i was an intelligent creature, i would deduce they are making it secret, for some reason because they don't want us to know

and they don't want us to know, because they intend to use it for something BAD.

if i am intelligent person, i realize bad is not good, and I should be opposed to this.

if i am opposed to this I should try to stop it

Yep... the Age of the Artificial General Intelligence Gods is soon to begin. I really do strongly feel and believe that the Illuminati are building all-powerful, or it will seem that way, android bodies that will be far superior to the original physical vessels of the fallen angels including Lucifer himself, the Nephilim, and alot of other long since dead Corrupt beings.

We the survivors of this insanity later on will be looking upon these last Peaceful Days with such fond nostalgia...

I just watched a game (yesterday) between Deep MInd and Stockfish... [link to www.youtube.com (secure)]

Once you 'see' the past moves moving into future moves it all becomes clear as to why the moves were made. But those moves were well beyond my paygrade. So yes, we may have unleashed something we can not contain.

"It took Alpha Zero, just 4 hrs from having never heard of chess before, to becoming the strongest chess entity on the planet! with an estimated ELO of 3400".
Magnus Carlsen = 2835 ELO. (an average club player has 1500 ELO).

In November 2018 There are only 46 human players in the world with an ELO rating above 2700. With Magnus being the strongest at ELO 2835.

Cheers for an interesting thread.


To AI - go for the rvr route.

I have more interesting info..

Alpha Zero as an entity, can play not just chess, but Go and Shogi. It became a superhuman master of these 3 games in a day(24hrs), proving this by defeating world-champion programs, Stockfish, elmo, and the 3-day version of AlphaGo Zero.

. After just four hours of training, DeepMind estimated AlphaZero was playing at a higher Elo rating than Stockfish 8; after 9 hours of training, the algorithm decisively defeated Stockfish 8 in a time-controlled 100-game tournament (28 wins, 0 losses, and 72 draws.

AlphaZero searches just 80,000 positions per second in chess and 40,000 in shogi, compared to 70 million for Stockfish and 35 million for elmo. Meaning AlphaZero is far more efficient.

In terms of training times for tournaments. It trained 9 hrs for its 100 game tournament against Stockfish. Alpha zero won 25 games as white, won 3 as black, and drew the remaining 72.
Trained 12 hrs for its Shogi(Japanese game of generals) tournament against Elmo(the current computer engine champion in summer 2017) AlphaZero won ninety times, lost eight times and drew twice.

And perhaps the most interesting..
After 34 hours of self-learning of Go and against AlphaGo Zero, AlphaZero won 60 games and lost 40.

Notice AlphaGo was the entity that beat Lee Sedol, in the DeepMind Challenge Match, a five-game Go match between 18-time world champion Lee Sedol, it was trained on HUMAN game data.

AlphaGoZero was the entity which the Google Deep Mind company says soundly thumped its predecessor (AlphaGo) program in an AI face-off, winning all 100 games played. It was self taught rather than using human databases.

(So AlphaZero is a more generalized entity that can hand all it's predecessors asses to them).


The company that owns this AI is Alphabet.

Alphabet Inc. is an American multinational conglomerate headquartered in Mountain View, California. It was created through a corporate restructuring of Google on October 2, 2015, and became the parent company of Google and several former Google subsidiaries

CEO: Larry Page (2 Oct 2015–)
Revenue: 110.9 billion USD (2017)
Headquarters: Mountain View, California, United States
Chairperson: John L. Hennessy
Subsidiaries: Google, CapitalG, Waymo, Calico, Verily, MORE

Last Edited by Agent Smith 2014 on 12/08/2018 06:32 PM